NVIDIA · Ada Lovelace · Data centre
NVIDIA L40S
An Ada Lovelace PCIe card for inference, fine-tuning, and graphics workloads, with FP8 support and no NVLink.
01 / SPECIFICATIONS
What the hardware holds
- Memory
- 48 GB
- Memory bandwidth
- 0.86 TB/s
- FP8 dense
- 733 TFLOPS
- FP16 / BF16 dense
- 362.05 TFLOPS
- Variant
- L40S
- Memory type
- GDDR6 (ECC)
- Form factor
- PCIe dual-slot, 4.4" x 10.5", passive
- Interconnect
- PCIe Gen4 x16 64 GB/s
- Max power
- 350 W
Source notes for L40S
Product page spec table lists FP16/BF16 Tensor '362.05 | 733*' and FP8 '733 | 1,466*' (* with sparsity); first value dense. Bandwidth 864 GB/s. Max power consumption 350 W. No NVLink. Peak rates based on GPU boost clock.
Compute figures are dense, without sparsity. NVIDIA source for L40S Checked 8 Nov 2026.
02 / MEMORY FIT
How many do you need to hold a model?
| Model size | FP16 / BF16 | FP8 | INT4 |
|---|---|---|---|
| 8B parameters | 1 GPU16 GB of weights | 1 GPU8 GB of weights | 1 GPU4 GB of weights |
| 70B parameters | 4 GPUs140 GB of weights | 2 GPUs70 GB of weights | 1 GPU35 GB of weights |
| 405B parameters | 21 GPUs810 GB of weights | 11 GPUs405 GB of weights | 6 GPUs203 GB of weights |
GPUs = parameters × bytes per parameter × 1.2 headroom ÷ 48 GB, rounded up. This covers weights only. Long contexts, large batches, and training need much more memory.
03 / OFFERS
Reviewed on-demand rates
| Provider | Configuration | Status | Source | ||
|---|---|---|---|---|---|
 Runpod | L40S1× L40S Pod · 48 GB | $1.09$1.09 ÷ 1 GPU | $1.091 GPU | Stock not verified | Rate card |
 DigitalOcean | L40S1× L40S on-demand Droplet · 48 GB | $1.57Published per GPU-hour | $1.571 GPU | Stock not verified | Rate card |
Per GPU-hour divides a machine's hourly price by its GPU count, so single GPUs and 8-GPU nodes compare fairly. The billed price is what one machine costs per hour. Storage, network transfer, and minimum terms can add to the bill.
04 / SELLERS
Who lists the L40S
- RunpodGPU cloud · Self-serve · Reviewed rate
CoreWeaveGPU cloud · Sales-led
NebiusGPU cloud · Self-serve
CrusoeGPU cloud · Self-serve
Vast.aiMarketplace · Self-serve- DigitalOceanGPU cloud · Self-serve · Reviewed rate
CoreWeave, Nebius, Crusoe, Vast.ai list the L40S without a rate we have reviewed. Ask them for a quote, especially for clusters of 8 or more GPUs.
05 / COMPARE WITH